Audio Source Separation Based on Convolutive Transfer Function and Frequency-Domain Lasso Optimization

Size: px
Start display at page:

Download "Audio Source Separation Based on Convolutive Transfer Function and Frequency-Domain Lasso Optimization"

Transcription

1 Audio Source Separation Based on Convolutive Transfer Function and Frequency-Domain Lasso Optimization Xiaofei Li, Laurent Girin, Radu Horaud To cite this version: Xiaofei Li, Laurent Girin, Radu Horaud. Audio Source Separation Based on Convolutive Transfer Function and Frequency-Domain Lasso Optimization. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Mar 2017, New Orleans, United States <hal > HAL Id: hal Submitted on 10 Jan 2017 HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

2 AUDIO SOURCE SEPARATION BASED ON CONVOLUTIVE TRANSFER FUNCTION AND FREQUENCY-DOMAIN LASSO OPTIMIZATION Xiaofei Li 1, Laurent Girin 1,2 and Radu Horaud 1 1 INRIA Grenoble Rhône-Alpes, 2 GIPSA-Lab & Univ. Grenoble Alpes ABSTRACT This paper addresses the problem of under-determined convolutive audio source separation in a semi-oracle configuration where the mixing filters are assumed to be known. We propose a separation procedure based on the convolutive transfer function (CTF), which is a more appropriate model for strongly reverberant signals than the widely-used multiplicative transfer function approximation. In the short-time Fourier transform domain, source signals are estimated by minimizing the mixture fitting cost using Lasso optimization, with a l 1 -norm regularization to exploit the spectral sparsity of source signals. Experiments show that the proposed method achieves satisfactory performance on highly reverberant speech mixtures, with a much lower computational cost compared to time-domain dual techniques. Index Terms Source separation, convolutive transfer function, l 1 -norm regularization 1. INTRODUCTION In this paper we address the problem of multichannel audio source separation (MASS) from, possibly underdetermined, convolutive mixtures (the number of sensors is lower than the number of sources). This problem is often divided into two subproblems, which are both difficult: identification of mixing filters and estimations of source signals. This paper focuses on audio source estimation assuming that the mixing filters are either known or an estimation is available. Most of convolutive source separation techniques are designed in the short time Fourier transform (STFT) domain. In this domain, the convolutive process is assumed to be well approximated at each TF bin by a product between the source STFT coefficient and the Fourier transform of the mixing filter. This assumption is called the multiplicative transfer function (MTF) approximation [1], or the narrowband approximation. Moreover, the sparsity of the audio signals in the timefrequency (TF) domain is a desirable effect. Based on these properties, the separation methods rely on binary masking of the mixture STFT bins, e.g. [2, 3], on the l 1 -norm minimiza- This research has received funding from the ERC Advanced Grand VHIA (#340113) and from EU-FP7 STREP project EARS (#609465). tion, e.g. [4], on probabilistic models for source signals, e.g. [5, 6], or on a combination of these methods. The MTF approximation is theoretically valid only if the length of the mixing filter impulse response is smaller than the length of the STFT window. In practice, this is very rarely the case, even for moderately reverberant environments, since the STFT window is limited to assume local stationarity of audio signals. Hence the MTF can be a poor approximation, fundamentally endangering the separation performance; this becomes critical for strongly reverberant environments. Yet, the MTF is poorly questioned in the MASS literature, and only a few studies attempted to tackle its limitations. In [6], the use of a full-rank spatial covariance matrix for the source images, instead of the rank-1 matrix corresponding to the MTF model [5], is claimed to overcome to some extent the limitations of MTF. A more direct approach to the problem is proposed in [7] where the source signals are estimated in the time domain by minimizing a wide-band l 2 -norm mixture-fitting cost, in which the exact source-filter convolution is used, using a Lasso optimization technique. [8] improved this wideband Lasso (W-Lasso) technique by a re-weighted scheme. The W-Lasso technique achieves quite good source separation performance in reverberant environments, at the price of a tremendous computation time. To represent convolution more accurately in the STFT domain, especially for the long filter case, cross-band filters (CBFs) were introduced in [9] in the context of linear system identification, as an alternative to MTF. Using the CBFs, an output STFT coefficient is represented as a summation over frequency bins of multiple convolutions between the input STFT coefficients and the TF-domain filter impulse response, along the frame coordinate. A convolutive transfer function (CTF) approximation is further introduced in [10] to simplify the analysis. Here, at each frequency, the output STFT coefficient is modeled as a (unique) convolution of the input STFT coefficients and the CTF, along the frame axis. The CBFs were recently considered for solving MASS [11], in combination with a high-resolution non-negative matrix factorization model of the source signal. A variational EM algorithm was proposed to estimate the filters and infer the source signals. Unfortunately, this method was observed to perform well only in a semi-blind setup where both filters and source parameters are initialized from the individual source images.

3 In this paper we propose to use the CTF for MASS quite differently than what was proposed in the past. Following the spirit of W-Lasso, the source signals are estimated, at each frequency, by minimizing a CTF-based l 2 -norm mixture fitting cost in the STFT domain. In addition, we add to the fitting cost a l 1 -norm regularizer such as to exploit the sparsity of TF-domain audio signals along the frames. By circumventing the MTF approximation, the proposed method achieves satisfactory source separation performance in reverberant environments, likewise W-Lasso, but since the STFT frame level is considered instead of the time-domain sample level, separation is obtained at a much lower computational burden, as shown by our experiments. Another potential advantage of the proposed method is that, compared to the timedomain mixing filters in W-Lasso, it may be easier to identify the CTF mixing filters from the mixture signals based on the TF-domain sparsity. The rest of this paper is organized as follows. Section 2 presents the CTF model. The proposed source separation method is given in Section 3. Experiments are presented in Section 4. Section 5 concludes the paper. 2. CONVOLUTIVE TRANSFER FUNCTION In a reverberant and noise-free environment, a source image y(n) is, in time domain, given by y(n) = a(n) s(n) (1) where s(n) and a(n) are the source signal and the impulse response of the propagating filter, respectively, and denotes linear convolution. With the usual MTF approximation, (1) is approximated in the STFT domain as y p,k = a k s p,k (2) where y p,k and s p,k are the STFT of the corresponding signals, and a k is the Fourier transform of the filter a(n), p [1, P ] is the frame index, N is the frame length, and k [0, N 1] is the frequency bin index. As discussed above, this approximation is only valid when the filter a(n) is shorter than the STFT window, which is often questionable. In this paper we therefore use the CTF model, i.e. y(n) is approximated in the STFT domain by: y p,k = p a p,ks p p,k = a p,k s p,k. (3) The filter CTF, i.e. the TF-domain impulse response a p,k, is related to the time-domain impulse response a(n) by: a p,k = (a(n) ζ k (n)) n=p L, (4) which represents the convolution with respect to the time index n evaluated at multiples of the frame step L, with ζ k (n) = e j 2π N kn + m= ω(m) ω(n + m), (5) where ω(n) and ω(n) denote the STFT analysis and synthesis windows, respectively. The CTF can be interpreted as: the time-domain convolution is transformed into a TF-domain convolution with a certain approximation error. 3. SEMI-BLIND SOURCE SEPARATION BASED ON CTF 3.1. Mixture model and source separation formulation We consider a multi-channel underdetermined convolutive mixture with J sources and I sensors (I < J). Based on the CTF model (3), in the STFT domain the sensor signals x i p,k, i [1, I] are given by: x i p,k = J y j p,k + ei p,k = j=1 J j=1 a i,j p,k sj p,k + ei p,k, (6) where a i,j p,k is the CTF from source j to sensor i, and ei p,k denotes the noise signal. For frequency k, let x C I P, s C J P and e C I P denote the matrices of sensor signals, source signals and noise signals, respectively, and let A C I J P denote the three-way CTF array. Since the proposed algorithm is frequency-wise, the frequency index k is omitted from now on. Then (6) can be written as: x = A s + e. (7) For underdetermined mixtures, the source signals cannot be estimated by inverting the filters. Instead, we estimate the source signals by minimizing a l 2 -norm mixture fitting cost. Moreover, the TF-domain sparsity of the speech sources is enforced using an l 1 -norm regularization term. Overall, the source separation is carried out by solving the following convex optimization problem: min A s x 2 s C J P 2 +λ s 1, (8) where the free parameter λ is chosen as a trade-off between the fitting term and the regularization term. Note that both the l 2 and l 1 norms on matrices in (8) are defined here as vector norms. As mentioned in the introduction, this problem was considered in [7] in the time-domain, as a wide-band (convolutive) version of the Lasso problem [12] or of the basis pursuit denoising problem [13]. [7] involved a large data size, leading to a low optimization convergence rate and a huge computational cost. Applying the convolutive Lasso to MASS in the STFT domain, thanks to the CTF model proposed here, enables to considerably reduce the computational burden and to directly exploit the STFT sparsity of audio signals Optimization using FISTA To solve the optimization problem (8), we adopt the fast iterative shrinkage-thresholding algorithm (FISTA) [14], as al-

4 ready done in [7] in the time-domain, with the notable difference that here the optimization process is carried on the complex domain. Let F(s) = A s x 2 2 denote the fitting cost function, which is L-Lipschitz differentiable. Its derivative is F(s) = à (A s x), (9) where the adjoint matrix à is obtained by conjugate transposing the source and channel indices, and then temporally reversing the filters. Similar to [7], the Lipschitz constant L is computed using the power iteration algorithm, which is summarized in Algorithm 1. In this work, v is initialized as a matrix composed of I replications of the first channel sensor signal. The convergence criterion consists in testing if the difference in amplitude of the inner product between the values of v, at the current and previous iterations, is larger than a threshold δ. The l 1 -norm regularization term in (8) is a lower semicontinuous function, non-differentiable at 0. The proximity operator of the function γ 1 at point z, aka the shrinkage operator, is defined as: Prox γ 1 (z) = argmin y 1 2 z y 2 2 +γ y 1. (10) The closed-form entrywise solution is given by: y i = z i z i max(0, z i γ). (11) Based on the derivative operator (9), on the Lipschitz constant L, and on the shrinkage operator, FISTA is summarized in Algorithm 2. s 0 is also initialized as the matrix composed of I replication of the first channel sensor signal. The convergence criterion is chosen as a threshold ɛ over the relative decrease of the objective cost in (8) at each iteration. Algorithm 1 Power iteration. Input: A, à Initialization: v C I P repeat w = à (A v) v = w/ w 2 until convergence L = w 2 4. EXPERIMENTAL EVALUATION To test the efficiency of the proposed source separation method, experiments were conducted with simulated binaural signals, under various acoustic conditions. Binaural room impulse responses (BRIR) were generated with the ROOMSIM simulator [15] and with the head related impulse Algorithm 2 FISTA Input: x, A, Ã, L Initialization: k = 1, s 0 C I P, z 0 = s 0, t 0 = 1 repeat 1. F(s k 1 ) = à (A s k 1 x) 2. s k = Prox (λ/l) 1 (z k 1 F(s k 1 )/L) 3. t k = (1 + (1 + 4t 2 k 1 ))/2 4. z k = s k + ((t k 1 1)/t k )(s k s k 1 ) 5. k = k + 1 until convergence response (HRIR) of a KEMAR dummy head [16]. The size of the room was 8 m 5 m 3 m. The KEMAR dummy head was located at (4 m, 1 m, 1.5 m). Speech signals from the TIMIT dataset [17] and then sampled at 16 khz were used as sources convolved with the simulated BRIRs to generate sensor signals. The speech sources were located at 1 m away from the dummy head with azimuth directions varying from 90 to 90, spaced by 5, and at an elevation of 0. Three reverberation times were tested, namely T 60 = 0.22 s, 0.5 s and 0.79 s. Moreover, the anechoic case was also tested. A set of underdetermined mixtures with 3, 4 and 5 sources were processed. For each experiment, 50 mixtures were generated. The STFT window was a Hamming window of 512 samples (32 ms), with 50% overlap. The free parameter λ was set at a fixed value of 10 3 through the whole set of experiments, as it was shown to be suitable for all the tested conditions. The power iteration convergence threshold δ was set to The FISTA convergence threshold was set to The CTF coefficients were computed from the known filter impulse responses using (4). For comparison, we tested four state-of-the-art source separation methods, all using in the same semi-blind configuration as the proposed method (i.e. with known mixing filters): the degenerate unmixing estimation technique (DUET) [2], which is based on time-frequency masking, assuming only a single source is active in each TF bin, the l 1 -norm minimization method (l 1 -MIN) [4], which assumes that at most I sources are active in each TF bin, the full-rank spatial covariance matrix (FR-SCM) method of [6], and the wide-band Lasso (W-Lasso) method of [7] with l 1 -norm regularization term on source STFT coefficients, and trade-off factor set to DUET and l 1 -MIN methods are based on the MTF approximation, i.e. the instantaneous mixing matrix in each frequency bin is employed. Since the BRIRs are longer than the STFT window, they have to be truncated to generate the mixing matrix with the Fourier transform. However we obtained better results using the Fourier transform of the HRIRs. For FR-SCM, the SCMs were individually estimated using each separate image source signal, following the line of the semi-oracle experiments in [6]. Then an EM was applied with the SCMs being kept fixed to the semi-oracle

5 Table 2: Source separation performance (SDR in db) for various number of sources (T 60 = 0.5 s). Number of sources Methods DUET [2] l 1 MIN [4] FR-SCM [6] W-Lasso [7] Proposed Fig. 1: Source separation performance (SDR) for 3-source mixtures as a function of reverberation time. Table 1: Computation times, in seconds, for a 3-source mixture and for various reverberation times. All the algorithms were implemented in Matlab. T 60 (s) Methods Anechoic DUET [2] l 1 MIN [4] FR-SCM [6] W-Lasso [7] Proposed values. Note that we did not compare our method with [11], since this method involves a specific rank-1 source model that is poorly appropriate for speech signals. The signal-to-distortion ratio (SDR) [18] in decibels (db), averaged over 50 mixtures for each condition, is used as the separation performance metric (only 5 mixtures are tested for W-Lasso due to its high computation time). Fig. 1 plots the SDR obtained for 3-source mixtures and for the 3 reverberation times. It can be seen in these plots that all five methods achieve high SDR in the anechoic case. DUET performs the worst, due to its limiting assumption that only a single source is active at each TF bin. l 1 -MIN and FR-SCM perform better by assuming that more sources can co-exist in a TF bin. W-Lasso achieves the highest SDR, thanks to the exact (time-domain) convolution model. The proposed method achieves a lower SDR than W-Lasso due to the CTF approximation error. As the reverberation time increases, the SDRs of DUET, of l 1 -MIN and to a least extent of FR-SCM, dramatically decrease. For DUET and for l 1 -MIN, the MTF approximation is no longer suitable when the filter impulse response is (much) longer than the STFT window. FR-SCM mitigates the problem by using a full-rank spatial covariance matrix, which models the reverberations, although to a limited extend. In contrast to these three methods, both W-Lasso and the proposed method achieve remarkably stable performances: the SDR actually increases with T 60, which is a bit surprising at first sight. This can be explained as follows: (i) the mixture models in W-Lasso and in the proposed method fit the actual mixture better than the MTF, and (ii) given a good fit of the mixture model, the longer the filter, the more information availalble to discriminate and separate different sources. Again, due to the CTF approximation error, the proposed method performs worse than W-Lasso by 2 to 5 db, depending on T 60. Table 1 shows the average computation time needed to process one mixture (averaged over the 50 test mixtures; the average mixture duration is about 4 s) for each method and each tested T 60. We can see in this table that the W-Lasso method is much more time-consuming than the others. DUET is the fastest. The computation time of FR-SCM is about 12% of the computation time of W-Lasso. The computation time of the proposed method and of l 1 -MIN is comparable, and is less than 1% of the computation time of W-Lasso. Finally, Table 2 displays the SDR for various number of sources, for T 60 = 0.5 s. As expected, the SDR of all five methods degrades when the source number increases. W- Lasso has the fastest degradation with increasing number of sources, whereas the proposed method seems more robust, so that for 5 sources, W-Lasso achieves only about 1.1 db SDR improvement over the proposed method. 5. CONCLUSION In this paper, we have proposed a semi-blind source separation method based on the CTF model and STFT-domain Lasso optimization. Overall, the proposed method and the timedomain W-Lasso perform prominently better than DUET, l 1 -MIN and FR-SCM. The proposed method drastically improves the computation efficiency over W-Lasso with an acceptable decrease in separation performance. It thus seems to be an excellent trade-off between the fast but inaccurate MTF model, and the exact but highly greedy time-domain W-Lasso, especially for filters that are much longer than the STFT window. Recently, it was proposed a CTF-based method to extract the direct path of a single source [19]. Based on the STFT-domain sparsity, the extension to multiple sources is under development, which altogether will enable the development of a CTF-based blind source separation method.

6 6. REFERENCES [1] Y. Avargel and I. Cohen, On multiplicative transfer function approximation in the short-time Fourier transform domain, IEEE Signal Processing Letters, vol. 14, no. 5, pp , [2] O. Yilmaz and S. Rickard, Blind separation of speech mixtures via time-frequency masking, IEEE Transactions on signal processing, vol. 52, no. 7, pp , [3] M. I. Mandel, R. J. Weiss, and D. Ellis, Model-based expectation-maximization source separation and localization, IEEE Transactions on Audio, Speech, and Language Processing, vol. 18, no. 2, pp , [4] S. Winter, W. Kellermann, H. Sawada, and S. Makino, Map-based underdetermined blind source separation of convolutive mixtures by hierarchical clustering and l 1 - norm minimization, EURASIP Journal on Applied Signal Processing, vol. 2007, no. 1, pp , [5] A. Ozerov and C. Févotte, Multichannel nonnegative matrix factorization in convolutive mixtures for audio source separation, IEEE Transactions on Audio, Speech, and Language Processing, vol. 18, no. 3, pp , [6] N. Duong, E. Vincent, and R. Gribonval, Underdetermined reverberant audio source separation using a full-rank spatial covariance model, IEEE Transactions on Audio, Speech, and Language Processing, vol. 18, no. 7, pp , [7] M. Kowalski, E. Vincent, and R. Gribonval, Beyond the narrowband approximation: Wideband convex methods for under-determined reverberant audio source separation, IEEE Transactions on Audio, Speech, and Language Processing, vol. 18, no. 7, pp , [8] S. Arberet, P. Vandergheynst, J-P. Carrillo, R. E.Thiran, and Y. Wiaux, Sparse reverberant audio source separation via reweighted analysis, IEEE Transactions on Audio, Speech, and Language Processing, vol. 21, no. 7, pp , [9] Y. Avargel and I. Cohen, System identification in the short-time Fourier transform domain with crossband filtering, IEEE Transactions on Audio, Speech, and Language Processing, vol. 15, no. 4, pp , [10] R. Talmon, I. Cohen, and S. Gannot, Relative transfer function identification using convolutive transfer function approximation, IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 17, no. 4, pp , [11] R. Badeau and M. D. Plumbley, Multichannel highresolution NMF for modeling convolutive mixtures of non-stationary signals in the time-frequency domain, IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 22, no. 11, pp , [12] R. Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society. Series B (Methodological), pp , [13] S. S. Chen, D. L. Donoho, and M. A. Saunders, Atomic decomposition by basis pursuit, SIAM review, vol. 43, no. 1, pp , [14] A. Beck and M. Teboulle, A fast iterative shrinkagethresholding algorithm for linear inverse problems, SIAM journal on imaging sciences, vol. 2, no. 1, pp , [15] D. R. Campbell, The ROOMSIM user guide (v3.3), [16] W. G. Gardner and K. D. Martin, HRTF measurements of a KEMAR dummy-head microphone, The Journal of the Acoustical Society of America, vol. 97, no. 6, pp , [17] J. S. Garofolo, L. F. Lamel, W. M. Fisher, J. G. Fiscus, D. S. Pallett, and N. L. Dahlgren, Getting started with the DARPA TIMIT CD-ROM: An acoustic phonetic continuous speech database, Tech. Rep. 107, National Institute of Standards and Technology, Gaithersburgh, MD, [18] E. Vincent, R. Gribonval, and C. Févotte, Performance measurement in blind audio source separation, IEEE transactions on audio, speech, and language processing, vol. 14, no. 4, pp , [19] X. Li, L. Girin, R. Horaud, and S. Gannot, Estimation of the direct-path relative transfer function for supervised sound-source localization, IEEE/ACM Transactions on Audio, Speech and Language Processing, vol. 24, no. 11, pp , 2016.

AN EM ALGORITHM FOR AUDIO SOURCE SEPARATION BASED ON THE CONVOLUTIVE TRANSFER FUNCTION

AN EM ALGORITHM FOR AUDIO SOURCE SEPARATION BASED ON THE CONVOLUTIVE TRANSFER FUNCTION AN EM ALGORITHM FOR AUDIO SOURCE SEPARATION BASED ON THE CONVOLUTIVE TRANSFER FUNCTION Xiaofei Li, 1 Laurent Girin, 2,1 Radu Horaud, 1 1 INRIA Grenoble Rhône-Alpes, France 2 Univ. Grenoble Alpes, GIPSA-Lab,

More information

Non-Stationary Noise Power Spectral Density Estimation Based on Regional Statistics

Non-Stationary Noise Power Spectral Density Estimation Based on Regional Statistics Non-Stationary Noise Power Spectral Density Estimation Based on Regional Statistics Xiaofei Li, Laurent Girin, Sharon Gannot, Radu Horaud To cite this version: Xiaofei Li, Laurent Girin, Sharon Gannot,

More information

ESTIMATION OF RELATIVE TRANSFER FUNCTION IN THE PRESENCE OF STATIONARY NOISE BASED ON SEGMENTAL POWER SPECTRAL DENSITY MATRIX SUBTRACTION

ESTIMATION OF RELATIVE TRANSFER FUNCTION IN THE PRESENCE OF STATIONARY NOISE BASED ON SEGMENTAL POWER SPECTRAL DENSITY MATRIX SUBTRACTION ESTIMATION OF RELATIVE TRANSFER FUNCTION IN THE PRESENCE OF STATIONARY NOISE BASED ON SEGMENTAL POWER SPECTRAL DENSITY MATRIX SUBTRACTION Xiaofei Li 1, Laurent Girin 1,, Radu Horaud 1 1 INRIA Grenoble

More information

Estimation of the Direct-Path Relative Transfer Function for Supervised Sound-Source Localization

Estimation of the Direct-Path Relative Transfer Function for Supervised Sound-Source Localization 1 Estimation of the Direct-Path Relative Transfer Function for Supervised Sound-Source Localization Xiaofei Li, Laurent Girin, Radu Horaud and Sharon Gannot arxiv:1509.03205v3 [cs.sd] 27 Jun 2016 Abstract

More information

Sparse filter models for solving permutation indeterminacy in convolutive blind source separation

Sparse filter models for solving permutation indeterminacy in convolutive blind source separation Sparse filter models for solving permutation indeterminacy in convolutive blind source separation Prasad Sudhakar, Rémi Gribonval To cite this version: Prasad Sudhakar, Rémi Gribonval. Sparse filter models

More information

A simple test to check the optimality of sparse signal approximations

A simple test to check the optimality of sparse signal approximations A simple test to check the optimality of sparse signal approximations Rémi Gribonval, Rosa Maria Figueras I Ventura, Pierre Vergheynst To cite this version: Rémi Gribonval, Rosa Maria Figueras I Ventura,

More information

Estimation of the Direct-Path Relative Transfer Function for Supervised Sound-Source Localization

Estimation of the Direct-Path Relative Transfer Function for Supervised Sound-Source Localization IEEE/ACM TRANSACTIONS ON AUDIO, SPEECH, AND LANGUAGE PROCESSING, VOL. 24, NO. 11, NOVEMBER 2016 2171 Estimation of the Direct-Path Relative Transfer Function for Supervised Sound-Source Localization Xiaofei

More information

SINGLE CHANNEL SPEECH MUSIC SEPARATION USING NONNEGATIVE MATRIX FACTORIZATION AND SPECTRAL MASKS. Emad M. Grais and Hakan Erdogan

SINGLE CHANNEL SPEECH MUSIC SEPARATION USING NONNEGATIVE MATRIX FACTORIZATION AND SPECTRAL MASKS. Emad M. Grais and Hakan Erdogan SINGLE CHANNEL SPEECH MUSIC SEPARATION USING NONNEGATIVE MATRIX FACTORIZATION AND SPECTRAL MASKS Emad M. Grais and Hakan Erdogan Faculty of Engineering and Natural Sciences, Sabanci University, Orhanli

More information

(3) where the mixing vector is the Fourier transform of are the STFT coefficients of the sources I. INTRODUCTION

(3) where the mixing vector is the Fourier transform of are the STFT coefficients of the sources I. INTRODUCTION 1830 IEEE TRANSACTIONS ON AUDIO, SPEECH, AND LANGUAGE PROCESSING, VOL. 18, NO. 7, SEPTEMBER 2010 Under-Determined Reverberant Audio Source Separation Using a Full-Rank Spatial Covariance Model Ngoc Q.

More information

A Variance Modeling Framework Based on Variational Autoencoders for Speech Enhancement

A Variance Modeling Framework Based on Variational Autoencoders for Speech Enhancement A Variance Modeling Framework Based on Variational Autoencoders for Speech Enhancement Simon Leglaive 1 Laurent Girin 1,2 Radu Horaud 1 1: Inria Grenoble Rhône-Alpes 2: Univ. Grenoble Alpes, Grenoble INP,

More information

Analysis of Boyer and Moore s MJRTY algorithm

Analysis of Boyer and Moore s MJRTY algorithm Analysis of Boyer and Moore s MJRTY algorithm Laurent Alonso, Edward M. Reingold To cite this version: Laurent Alonso, Edward M. Reingold. Analysis of Boyer and Moore s MJRTY algorithm. Information Processing

More information

Rectified binaural ratio: A complex T-distributed feature for robust sound localization

Rectified binaural ratio: A complex T-distributed feature for robust sound localization Rectified binaural ratio: A complex T-distributed feature for robust sound localization Antoine Deleforge, Florence Forbes To cite this version: Antoine Deleforge, Florence Forbes. Rectified binaural ratio:

More information

An Inverse-Gamma Source Variance Prior with Factorized Parameterization for Audio Source Separation

An Inverse-Gamma Source Variance Prior with Factorized Parameterization for Audio Source Separation An Inverse-Gamma Source Variance Prior with Factorized Parameterization or Audio Source Separation Dionyssos Kounades-Bastian, Laurent Girin, Xavier Alameda-Pineda, Sharon Gannot, Radu Horaud To cite this

More information

Sound intensity as a function of sound insulation partition

Sound intensity as a function of sound insulation partition Sound intensity as a function of sound insulation partition S. Cvetkovic, R. Prascevic To cite this version: S. Cvetkovic, R. Prascevic. Sound intensity as a function of sound insulation partition. Journal

More information

Sensitivity of hybrid filter banks A/D converters to analog realization errors and finite word length

Sensitivity of hybrid filter banks A/D converters to analog realization errors and finite word length Sensitivity of hybrid filter banks A/D converters to analog realization errors and finite word length Tudor Petrescu, Jacques Oksman To cite this version: Tudor Petrescu, Jacques Oksman. Sensitivity of

More information

Scalable audio separation with light Kernel Additive Modelling

Scalable audio separation with light Kernel Additive Modelling Scalable audio separation with light Kernel Additive Modelling Antoine Liutkus 1, Derry Fitzgerald 2, Zafar Rafii 3 1 Inria, Université de Lorraine, LORIA, UMR 7503, France 2 NIMBUS Centre, Cork Institute

More information

Vibro-acoustic simulation of a car window

Vibro-acoustic simulation of a car window Vibro-acoustic simulation of a car window Christophe Barras To cite this version: Christophe Barras. Vibro-acoustic simulation of a car window. Société Française d Acoustique. Acoustics 12, Apr 12, Nantes,

More information

Speech enhancement based on nonnegative matrix factorization with mixed group sparsity constraint

Speech enhancement based on nonnegative matrix factorization with mixed group sparsity constraint Speech enhancement based on nonnegative matrix factorization with mixed group sparsity constraint Hien-Thanh Duong, Quoc-Cuong Nguyen, Cong-Phuong Nguyen, Thanh-Huan Tran, Ngoc Q. K. Duong To cite this

More information

A Probability Model for Interaural Phase Difference

A Probability Model for Interaural Phase Difference A Probability Model for Interaural Phase Difference Michael I. Mandel, Daniel P.W. Ellis Department of Electrical Engineering Columbia University, New York, New York {mim,dpwe}@ee.columbia.edu Abstract

More information

Methylation-associated PHOX2B gene silencing is a rare event in human neuroblastoma.

Methylation-associated PHOX2B gene silencing is a rare event in human neuroblastoma. Methylation-associated PHOX2B gene silencing is a rare event in human neuroblastoma. Loïc De Pontual, Delphine Trochet, Franck Bourdeaut, Sophie Thomas, Heather Etchevers, Agnes Chompret, Véronique Minard,

More information

Covariance smoothing and consistent Wiener filtering for artifact reduction in audio source separation

Covariance smoothing and consistent Wiener filtering for artifact reduction in audio source separation Covariance smoothing and consistent Wiener filtering for artifact reduction in audio source separation Emmanuel Vincent METISS Team Inria Rennes - Bretagne Atlantique E. Vincent (Inria) Artifact reduction

More information

Sparsity Measure and the Detection of Significant Data

Sparsity Measure and the Detection of Significant Data Sparsity Measure and the Detection of Significant Data Abdourrahmane Atto, Dominique Pastor, Grégoire Mercier To cite this version: Abdourrahmane Atto, Dominique Pastor, Grégoire Mercier. Sparsity Measure

More information

A new simple recursive algorithm for finding prime numbers using Rosser s theorem

A new simple recursive algorithm for finding prime numbers using Rosser s theorem A new simple recursive algorithm for finding prime numbers using Rosser s theorem Rédoane Daoudi To cite this version: Rédoane Daoudi. A new simple recursive algorithm for finding prime numbers using Rosser

More information

Acoustic Vector Sensor based Speech Source Separation with Mixed Gaussian-Laplacian Distributions

Acoustic Vector Sensor based Speech Source Separation with Mixed Gaussian-Laplacian Distributions Acoustic Vector Sensor based Speech Source Separation with Mixed Gaussian-Laplacian Distributions Xiaoyi Chen, Atiyeh Alinaghi, Xionghu Zhong and Wenwu Wang Department of Acoustic Engineering, School of

More information

Blind Spectral-GMM Estimation for Underdetermined Instantaneous Audio Source Separation

Blind Spectral-GMM Estimation for Underdetermined Instantaneous Audio Source Separation Blind Spectral-GMM Estimation for Underdetermined Instantaneous Audio Source Separation Simon Arberet 1, Alexey Ozerov 2, Rémi Gribonval 1, and Frédéric Bimbot 1 1 METISS Group, IRISA-INRIA Campus de Beaulieu,

More information

Source counting in real-time sound source localization using a circular microphone array

Source counting in real-time sound source localization using a circular microphone array Source counting in real-time sound source localization using a circular microphone array Despoina Pavlidi, Anthony Griffin, Matthieu Puigt, Athanasios Mouchtaris To cite this version: Despoina Pavlidi,

More information

Learning an Adaptive Dictionary Structure for Efficient Image Sparse Coding

Learning an Adaptive Dictionary Structure for Efficient Image Sparse Coding Learning an Adaptive Dictionary Structure for Efficient Image Sparse Coding Jérémy Aghaei Mazaheri, Christine Guillemot, Claude Labit To cite this version: Jérémy Aghaei Mazaheri, Christine Guillemot,

More information

Tropical Graph Signal Processing

Tropical Graph Signal Processing Tropical Graph Signal Processing Vincent Gripon To cite this version: Vincent Gripon. Tropical Graph Signal Processing. 2017. HAL Id: hal-01527695 https://hal.archives-ouvertes.fr/hal-01527695v2

More information

Easter bracelets for years

Easter bracelets for years Easter bracelets for 5700000 years Denis Roegel To cite this version: Denis Roegel. Easter bracelets for 5700000 years. [Research Report] 2014. HAL Id: hal-01009457 https://hal.inria.fr/hal-01009457

More information

Accounting for Missing Data in Sparse Wavelet Representation of Observation Error Correlations

Accounting for Missing Data in Sparse Wavelet Representation of Observation Error Correlations Accounting for Missing Data in Sparse Wavelet Representation of Observation Error Correlations Vincent Chabot, Maëlle Nodet, Arthur Vidard To cite this version: Vincent Chabot, Maëlle Nodet, Arthur Vidard.

More information

On Symmetric Norm Inequalities And Hermitian Block-Matrices

On Symmetric Norm Inequalities And Hermitian Block-Matrices On Symmetric Norm Inequalities And Hermitian lock-matrices Antoine Mhanna To cite this version: Antoine Mhanna On Symmetric Norm Inequalities And Hermitian lock-matrices 015 HAL Id: hal-0131860

More information

Underdetermined Instantaneous Audio Source Separation via Local Gaussian Modeling

Underdetermined Instantaneous Audio Source Separation via Local Gaussian Modeling Underdetermined Instantaneous Audio Source Separation via Local Gaussian Modeling Emmanuel Vincent, Simon Arberet, and Rémi Gribonval METISS Group, IRISA-INRIA Campus de Beaulieu, 35042 Rennes Cedex, France

More information

Hardware Operator for Simultaneous Sine and Cosine Evaluation

Hardware Operator for Simultaneous Sine and Cosine Evaluation Hardware Operator for Simultaneous Sine and Cosine Evaluation Arnaud Tisserand To cite this version: Arnaud Tisserand. Hardware Operator for Simultaneous Sine and Cosine Evaluation. ICASSP 6: International

More information

Dissipative Systems Analysis and Control, Theory and Applications: Addendum/Erratum

Dissipative Systems Analysis and Control, Theory and Applications: Addendum/Erratum Dissipative Systems Analysis and Control, Theory and Applications: Addendum/Erratum Bernard Brogliato To cite this version: Bernard Brogliato. Dissipative Systems Analysis and Control, Theory and Applications:

More information

Informed Audio Source Separation: A Comparative Study

Informed Audio Source Separation: A Comparative Study Informed Audio Source Separation: A Comparative Study Antoine Liutkus, Stanislaw Gorlow, Nicolas Sturmel, Shuhua Zhang, Laurent Girin, Roland Badeau, Laurent Daudet, Sylvain Marchand, Gaël Richard To cite

More information

On Newton-Raphson iteration for multiplicative inverses modulo prime powers

On Newton-Raphson iteration for multiplicative inverses modulo prime powers On Newton-Raphson iteration for multiplicative inverses modulo prime powers Jean-Guillaume Dumas To cite this version: Jean-Guillaume Dumas. On Newton-Raphson iteration for multiplicative inverses modulo

More information

On Symmetric Norm Inequalities And Hermitian Block-Matrices

On Symmetric Norm Inequalities And Hermitian Block-Matrices On Symmetric Norm Inequalities And Hermitian lock-matrices Antoine Mhanna To cite this version: Antoine Mhanna On Symmetric Norm Inequalities And Hermitian lock-matrices 016 HAL Id: hal-0131860

More information

A Novel Aggregation Method based on Graph Matching for Algebraic MultiGrid Preconditioning of Sparse Linear Systems

A Novel Aggregation Method based on Graph Matching for Algebraic MultiGrid Preconditioning of Sparse Linear Systems A Novel Aggregation Method based on Graph Matching for Algebraic MultiGrid Preconditioning of Sparse Linear Systems Pasqua D Ambra, Alfredo Buttari, Daniela Di Serafino, Salvatore Filippone, Simone Gentile,

More information

Completeness of the Tree System for Propositional Classical Logic

Completeness of the Tree System for Propositional Classical Logic Completeness of the Tree System for Propositional Classical Logic Shahid Rahman To cite this version: Shahid Rahman. Completeness of the Tree System for Propositional Classical Logic. Licence. France.

More information

Dispersion relation results for VCS at JLab

Dispersion relation results for VCS at JLab Dispersion relation results for VCS at JLab G. Laveissiere To cite this version: G. Laveissiere. Dispersion relation results for VCS at JLab. Compton Scattering from Low to High Momentum Transfer, Mar

More information

Case report on the article Water nanoelectrolysis: A simple model, Journal of Applied Physics (2017) 122,

Case report on the article Water nanoelectrolysis: A simple model, Journal of Applied Physics (2017) 122, Case report on the article Water nanoelectrolysis: A simple model, Journal of Applied Physics (2017) 122, 244902 Juan Olives, Zoubida Hammadi, Roger Morin, Laurent Lapena To cite this version: Juan Olives,

More information

Temporal and time-frequency correlation-based blind source separation methods. Part I : Determined and underdetermined linear instantaneous mixtures

Temporal and time-frequency correlation-based blind source separation methods. Part I : Determined and underdetermined linear instantaneous mixtures Temporal and time-frequency correlation-based blind source separation methods. Part I : Determined and underdetermined linear instantaneous mixtures Yannick Deville, Matthieu Puigt To cite this version:

More information

Linear Quadratic Zero-Sum Two-Person Differential Games

Linear Quadratic Zero-Sum Two-Person Differential Games Linear Quadratic Zero-Sum Two-Person Differential Games Pierre Bernhard To cite this version: Pierre Bernhard. Linear Quadratic Zero-Sum Two-Person Differential Games. Encyclopaedia of Systems and Control,

More information

A proximal approach to the inversion of ill-conditioned matrices

A proximal approach to the inversion of ill-conditioned matrices A proximal approach to the inversion of ill-conditioned matrices Pierre Maréchal, Aude Rondepierre To cite this version: Pierre Maréchal, Aude Rondepierre. A proximal approach to the inversion of ill-conditioned

More information

Physics-driven structured cosparse modeling for source localization

Physics-driven structured cosparse modeling for source localization Physics-driven structured cosparse modeling for source localization Sangnam Nam, Rémi Gribonval To cite this version: Sangnam Nam, Rémi Gribonval. Physics-driven structured cosparse modeling for source

More information

Towards an active anechoic room

Towards an active anechoic room Towards an active anechoic room Dominique Habault, Philippe Herzog, Emmanuel Friot, Cédric Pinhède To cite this version: Dominique Habault, Philippe Herzog, Emmanuel Friot, Cédric Pinhède. Towards an active

More information

Classification of high dimensional data: High Dimensional Discriminant Analysis

Classification of high dimensional data: High Dimensional Discriminant Analysis Classification of high dimensional data: High Dimensional Discriminant Analysis Charles Bouveyron, Stephane Girard, Cordelia Schmid To cite this version: Charles Bouveyron, Stephane Girard, Cordelia Schmid.

More information

COLLABORATIVE SPEECH DEREVERBERATION: REGULARIZED TENSOR FACTORIZATION FOR CROWDSOURCED MULTI-CHANNEL RECORDINGS. Sanna Wager, Minje Kim

COLLABORATIVE SPEECH DEREVERBERATION: REGULARIZED TENSOR FACTORIZATION FOR CROWDSOURCED MULTI-CHANNEL RECORDINGS. Sanna Wager, Minje Kim COLLABORATIVE SPEECH DEREVERBERATION: REGULARIZED TENSOR FACTORIZATION FOR CROWDSOURCED MULTI-CHANNEL RECORDINGS Sanna Wager, Minje Kim Indiana University School of Informatics, Computing, and Engineering

More information

A non-commutative algorithm for multiplying (7 7) matrices using 250 multiplications

A non-commutative algorithm for multiplying (7 7) matrices using 250 multiplications A non-commutative algorithm for multiplying (7 7) matrices using 250 multiplications Alexandre Sedoglavic To cite this version: Alexandre Sedoglavic. A non-commutative algorithm for multiplying (7 7) matrices

More information

Expectation-maximization algorithms for Itakura-Saito nonnegative matrix factorization

Expectation-maximization algorithms for Itakura-Saito nonnegative matrix factorization Expectation-maximization algorithms for Itakura-Saito nonnegative matrix factorization Paul Magron, Tuomas Virtanen To cite this version: Paul Magron, Tuomas Virtanen. Expectation-maximization algorithms

More information

LONG-TERM REVERBERATION MODELING FOR UNDER-DETERMINED AUDIO SOURCE SEPARATION WITH APPLICATION TO VOCAL MELODY EXTRACTION.

LONG-TERM REVERBERATION MODELING FOR UNDER-DETERMINED AUDIO SOURCE SEPARATION WITH APPLICATION TO VOCAL MELODY EXTRACTION. LONG-TERM REVERBERATION MODELING FOR UNDER-DETERMINED AUDIO SOURCE SEPARATION WITH APPLICATION TO VOCAL MELODY EXTRACTION. Romain Hennequin Deezer R&D 10 rue d Athènes, 75009 Paris, France rhennequin@deezer.com

More information

The constrained stochastic matched filter subspace tracking

The constrained stochastic matched filter subspace tracking The constrained stochastic matched filter subspace tracking Maissa Chagmani, Bernard Xerri, Bruno Borloz, Claude Jauffret To cite this version: Maissa Chagmani, Bernard Xerri, Bruno Borloz, Claude Jauffret.

More information

REAL-TIME TIME-FREQUENCY BASED BLIND SOURCE SEPARATION. Scott Rickard, Radu Balan, Justinian Rosca. Siemens Corporate Research Princeton, NJ 08540

REAL-TIME TIME-FREQUENCY BASED BLIND SOURCE SEPARATION. Scott Rickard, Radu Balan, Justinian Rosca. Siemens Corporate Research Princeton, NJ 08540 REAL-TIME TIME-FREQUENCY BASED BLIND SOURCE SEPARATION Scott Rickard, Radu Balan, Justinian Rosca Siemens Corporate Research Princeton, NJ 84 fscott.rickard,radu.balan,justinian.roscag@scr.siemens.com

More information

Quantum efficiency and metastable lifetime measurements in ruby ( Cr 3+ : Al2O3) via lock-in rate-window photothermal radiometry

Quantum efficiency and metastable lifetime measurements in ruby ( Cr 3+ : Al2O3) via lock-in rate-window photothermal radiometry Quantum efficiency and metastable lifetime measurements in ruby ( Cr 3+ : Al2O3) via lock-in rate-window photothermal radiometry A. Mandelis, Z. Chen, R. Bleiss To cite this version: A. Mandelis, Z. Chen,

More information

Automatic detection of the number of Raypaths

Automatic detection of the number of Raypaths Automatic detection of the number of Raypaths Longyu Jiang, Jerome Mars To cite this version: Longyu Jiang, Jerome Mars. Automatic detection of the number of Raypaths. OCEANS MTS/IEEE Kona - Oceans of

More information

Passerelle entre les arts : la sculpture sonore

Passerelle entre les arts : la sculpture sonore Passerelle entre les arts : la sculpture sonore Anaïs Rolez To cite this version: Anaïs Rolez. Passerelle entre les arts : la sculpture sonore. Article destiné à l origine à la Revue de l Institut National

More information

Simulation and measurement of loudspeaker nonlinearity with a broad-band noise excitation

Simulation and measurement of loudspeaker nonlinearity with a broad-band noise excitation Simulation and measurement of loudspeaker nonlinearity with a broad-band noise excitation Andrzej Dobrucki, Rafal Siczek To cite this version: Andrzej Dobrucki, Rafal Siczek. Simulation and measurement

More information

A Slice Based 3-D Schur-Cohn Stability Criterion

A Slice Based 3-D Schur-Cohn Stability Criterion A Slice Based 3-D Schur-Cohn Stability Criterion Ioana Serban, Mohamed Najim To cite this version: Ioana Serban, Mohamed Najim. A Slice Based 3-D Schur-Cohn Stability Criterion. ICASSP 007, Apr 007, Honolulu,

More information

Multiple sensor fault detection in heat exchanger system

Multiple sensor fault detection in heat exchanger system Multiple sensor fault detection in heat exchanger system Abdel Aïtouche, Didier Maquin, Frédéric Busson To cite this version: Abdel Aïtouche, Didier Maquin, Frédéric Busson. Multiple sensor fault detection

More information

On size, radius and minimum degree

On size, radius and minimum degree On size, radius and minimum degree Simon Mukwembi To cite this version: Simon Mukwembi. On size, radius and minimum degree. Discrete Mathematics and Theoretical Computer Science, DMTCS, 2014, Vol. 16 no.

More information

Full-order observers for linear systems with unknown inputs

Full-order observers for linear systems with unknown inputs Full-order observers for linear systems with unknown inputs Mohamed Darouach, Michel Zasadzinski, Shi Jie Xu To cite this version: Mohamed Darouach, Michel Zasadzinski, Shi Jie Xu. Full-order observers

More information

Exogenous input estimation in Electronic Power Steering (EPS) systems

Exogenous input estimation in Electronic Power Steering (EPS) systems Exogenous input estimation in Electronic Power Steering (EPS) systems Valentina Ciarla, Carlos Canudas de Wit, Franck Quaine, Violaine Cahouet To cite this version: Valentina Ciarla, Carlos Canudas de

More information

Fast Computation of Moore-Penrose Inverse Matrices

Fast Computation of Moore-Penrose Inverse Matrices Fast Computation of Moore-Penrose Inverse Matrices Pierre Courrieu To cite this version: Pierre Courrieu. Fast Computation of Moore-Penrose Inverse Matrices. Neural Information Processing - Letters and

More information

Enhancing Fetal ECG Using Gaussian Process

Enhancing Fetal ECG Using Gaussian Process Enhancing Fetal ECG Using Gaussian Process Saman Noorzadeh, Bertrand Rivet, Pierre-Yves Guméry To cite this version: Saman Noorzadeh, Bertrand Rivet, Pierre-Yves Guméry. Enhancing Fetal ECG Using Gaussian

More information

Decentralized Interference Channels with Noisy Feedback Possess Pareto Optimal Nash Equilibria

Decentralized Interference Channels with Noisy Feedback Possess Pareto Optimal Nash Equilibria Decentralized Interference Channels with Noisy Feedback Possess Pareto Optimal Nash Equilibria Samir M. Perlaza Ravi Tandon H. Vincent Poor To cite this version: Samir M. Perlaza Ravi Tandon H. Vincent

More information

Multiple Sound Source Counting and Localization Based on Spatial Principal Eigenvector

Multiple Sound Source Counting and Localization Based on Spatial Principal Eigenvector INTERSPEECH 27 August 2 24, 27, Stockholm, Sweden Multiple Sound Source Counting and Localization Based on Spatial Principal Eigenvector Bing Yang, Hong Liu, Cheng Pang Key Laboratory of Machine Perception,

More information

NOISE ROBUST RELATIVE TRANSFER FUNCTION ESTIMATION. M. Schwab, P. Noll, and T. Sikora. Technical University Berlin, Germany Communication System Group

NOISE ROBUST RELATIVE TRANSFER FUNCTION ESTIMATION. M. Schwab, P. Noll, and T. Sikora. Technical University Berlin, Germany Communication System Group NOISE ROBUST RELATIVE TRANSFER FUNCTION ESTIMATION M. Schwab, P. Noll, and T. Sikora Technical University Berlin, Germany Communication System Group Einsteinufer 17, 1557 Berlin (Germany) {schwab noll

More information

L institution sportive : rêve et illusion

L institution sportive : rêve et illusion L institution sportive : rêve et illusion Hafsi Bedhioufi, Sida Ayachi, Imen Ben Amar To cite this version: Hafsi Bedhioufi, Sida Ayachi, Imen Ben Amar. L institution sportive : rêve et illusion. Revue

More information

Determination of absorption characteristic of materials on basis of sound intensity measurement

Determination of absorption characteristic of materials on basis of sound intensity measurement Determination of absorption characteristic of materials on basis of sound intensity measurement R. Prascevic, A. Milosevic, S. Cvetkovic To cite this version: R. Prascevic, A. Milosevic, S. Cvetkovic.

More information

Widely Linear Estimation with Complex Data

Widely Linear Estimation with Complex Data Widely Linear Estimation with Complex Data Bernard Picinbono, Pascal Chevalier To cite this version: Bernard Picinbono, Pascal Chevalier. Widely Linear Estimation with Complex Data. IEEE Transactions on

More information

Analyzing large-scale spike trains data with spatio-temporal constraints

Analyzing large-scale spike trains data with spatio-temporal constraints Analyzing large-scale spike trains data with spatio-temporal constraints Hassan Nasser, Olivier Marre, Selim Kraria, Thierry Viéville, Bruno Cessac To cite this version: Hassan Nasser, Olivier Marre, Selim

More information

Smart Bolometer: Toward Monolithic Bolometer with Smart Functions

Smart Bolometer: Toward Monolithic Bolometer with Smart Functions Smart Bolometer: Toward Monolithic Bolometer with Smart Functions Matthieu Denoual, Gilles Allègre, Patrick Attia, Olivier De Sagazan To cite this version: Matthieu Denoual, Gilles Allègre, Patrick Attia,

More information

On Poincare-Wirtinger inequalities in spaces of functions of bounded variation

On Poincare-Wirtinger inequalities in spaces of functions of bounded variation On Poincare-Wirtinger inequalities in spaces of functions of bounded variation Maïtine Bergounioux To cite this version: Maïtine Bergounioux. On Poincare-Wirtinger inequalities in spaces of functions of

More information

Evolution of the cooperation and consequences of a decrease in plant diversity on the root symbiont diversity

Evolution of the cooperation and consequences of a decrease in plant diversity on the root symbiont diversity Evolution of the cooperation and consequences of a decrease in plant diversity on the root symbiont diversity Marie Duhamel To cite this version: Marie Duhamel. Evolution of the cooperation and consequences

More information

Music transcription with ISA and HMM

Music transcription with ISA and HMM Music transcription with ISA and HMM Emmanuel Vincent, Xavier Rodet To cite this version: Emmanuel Vincent, Xavier Rodet. Music transcription with ISA and HMM. 5th Int. Conf. on Independent Component Analysis

More information

Thomas Lugand. To cite this version: HAL Id: tel

Thomas Lugand. To cite this version: HAL Id: tel Contribution à la Modélisation et à l Optimisation de la Machine Asynchrone Double Alimentation pour des Applications Hydrauliques de Pompage Turbinage Thomas Lugand To cite this version: Thomas Lugand.

More information

Entropy-constrained quantization of exponentially damped sinusoids parameters

Entropy-constrained quantization of exponentially damped sinusoids parameters Entropy-constrained quantization of exponentially damped sinusoids parameters Olivier Derrien, Roland Badeau, Gaël Richard To cite this version: Olivier Derrien, Roland Badeau, Gaël Richard. Entropy-constrained

More information

Two-step centered spatio-temporal auto-logistic regression model

Two-step centered spatio-temporal auto-logistic regression model Two-step centered spatio-temporal auto-logistic regression model Anne Gégout-Petit, Shuxian Li To cite this version: Anne Gégout-Petit, Shuxian Li. Two-step centered spatio-temporal auto-logistic regression

More information

Matrix Factorization for Speech Enhancement

Matrix Factorization for Speech Enhancement Matrix Factorization for Speech Enhancement Peter Li Peter.Li@nyu.edu Yijun Xiao ryjxiao@nyu.edu 1 Introduction In this report, we explore techniques for speech enhancement using matrix factorization.

More information

A diagonal plus low-rank covariance model for computationally efficient source separation

A diagonal plus low-rank covariance model for computationally efficient source separation A diagonal plus low-ran covariance model for computationally efficient source separation Antoine Liutus, Kazuyoshi Yoshii To cite this version: Antoine Liutus, Kazuyoshi Yoshii. A diagonal plus low-ran

More information

Audio Source Separation With a Single Sensor

Audio Source Separation With a Single Sensor Audio Source Separation With a Single Sensor Laurent Benaroya, Frédéric Bimbot, Rémi Gribonval To cite this version: Laurent Benaroya, Frédéric Bimbot, Rémi Gribonval. Audio Source Separation With a Single

More information

A novel method for estimating the flicker level generated by a wave energy farm composed of devices operated in variable speed mode

A novel method for estimating the flicker level generated by a wave energy farm composed of devices operated in variable speed mode A novel method for estimating the flicker level generated by a wave energy farm composed of devices operated in variable speed mode Anne Blavette, Dara O Sullivan, Ray Alcorn, Mohamed Machmoum, Michael

More information

A Simple Proof of P versus NP

A Simple Proof of P versus NP A Simple Proof of P versus NP Frank Vega To cite this version: Frank Vega. A Simple Proof of P versus NP. 2016. HAL Id: hal-01281254 https://hal.archives-ouvertes.fr/hal-01281254 Submitted

More information

Impulse response measurement of ultrasonic transducers

Impulse response measurement of ultrasonic transducers Impulse response measurement of ultrasonic transducers F. Kadlec To cite this version: F. Kadlec. Impulse response measurement of ultrasonic transducers. Journal de Physique IV Colloque, 1994, 04 (C5),

More information

Factorisation of RSA-704 with CADO-NFS

Factorisation of RSA-704 with CADO-NFS Factorisation of RSA-704 with CADO-NFS Shi Bai, Emmanuel Thomé, Paul Zimmermann To cite this version: Shi Bai, Emmanuel Thomé, Paul Zimmermann. Factorisation of RSA-704 with CADO-NFS. 2012. HAL Id: hal-00760322

More information

NMF WITH SPECTRAL AND TEMPORAL CONTINUITY CRITERIA FOR MONAURAL SOUND SOURCE SEPARATION. Julian M. Becker, Christian Sohn and Christian Rohlfing

NMF WITH SPECTRAL AND TEMPORAL CONTINUITY CRITERIA FOR MONAURAL SOUND SOURCE SEPARATION. Julian M. Becker, Christian Sohn and Christian Rohlfing NMF WITH SPECTRAL AND TEMPORAL CONTINUITY CRITERIA FOR MONAURAL SOUND SOURCE SEPARATION Julian M. ecker, Christian Sohn Christian Rohlfing Institut für Nachrichtentechnik RWTH Aachen University D-52056

More information

From Unstructured 3D Point Clouds to Structured Knowledge - A Semantics Approach

From Unstructured 3D Point Clouds to Structured Knowledge - A Semantics Approach From Unstructured 3D Point Clouds to Structured Knowledge - A Semantics Approach Christophe Cruz, Helmi Ben Hmida, Frank Boochs, Christophe Nicolle To cite this version: Christophe Cruz, Helmi Ben Hmida,

More information

A non-commutative algorithm for multiplying (7 7) matrices using 250 multiplications

A non-commutative algorithm for multiplying (7 7) matrices using 250 multiplications A non-commutative algorithm for multiplying (7 7) matrices using 250 multiplications Alexandre Sedoglavic To cite this version: Alexandre Sedoglavic. A non-commutative algorithm for multiplying (7 7) matrices

More information

Finite element computation of leaky modes in straight and helical elastic waveguides

Finite element computation of leaky modes in straight and helical elastic waveguides Finite element computation of leaky modes in straight and helical elastic waveguides Khac-Long Nguyen, Fabien Treyssede, Christophe Hazard, Anne-Sophie Bonnet-Ben Dhia To cite this version: Khac-Long Nguyen,

More information

A Conjugate Gradient Algorithm for Blind Sensor Calibration in Sparse Recovery

A Conjugate Gradient Algorithm for Blind Sensor Calibration in Sparse Recovery A Conjugate Gradient Algorithm for Blind Sensor Calibration in Sparse Recovery Hao Shen, Martin Kleinsteuber, Cagdas Bilen, Rémi Gribonval To cite this version: Hao Shen, Martin Kleinsteuber, Cagdas Bilen,

More information

Soundness of the System of Semantic Trees for Classical Logic based on Fitting and Smullyan

Soundness of the System of Semantic Trees for Classical Logic based on Fitting and Smullyan Soundness of the System of Semantic Trees for Classical Logic based on Fitting and Smullyan Shahid Rahman To cite this version: Shahid Rahman. Soundness of the System of Semantic Trees for Classical Logic

More information

Generalized Constraints for NMF with Application to Informed Source Separation

Generalized Constraints for NMF with Application to Informed Source Separation Generalized Constraints for with Application to Informed Source Separation Christian Rohlfing and Julian M. Becker Institut für Nachrichtentechnik RWTH Aachen University, D2056 Aachen, Germany Email: rohlfing@ient.rwth-aachen.de

More information

SOLAR RADIATION ESTIMATION AND PREDICTION USING MEASURED AND PREDICTED AEROSOL OPTICAL DEPTH

SOLAR RADIATION ESTIMATION AND PREDICTION USING MEASURED AND PREDICTED AEROSOL OPTICAL DEPTH SOLAR RADIATION ESTIMATION AND PREDICTION USING MEASURED AND PREDICTED AEROSOL OPTICAL DEPTH Carlos M. Fernández-Peruchena, Martín Gastón, Maria V Guisado, Ana Bernardos, Íñigo Pagola, Lourdes Ramírez

More information

Two Dimensional Linear Phase Multiband Chebyshev FIR Filter

Two Dimensional Linear Phase Multiband Chebyshev FIR Filter Two Dimensional Linear Phase Multiband Chebyshev FIR Filter Vinay Kumar, Bhooshan Sunil To cite this version: Vinay Kumar, Bhooshan Sunil. Two Dimensional Linear Phase Multiband Chebyshev FIR Filter. Acta

More information

An EM Algorithm for Localizing Multiple Sound Sources in Reverberant Environments

An EM Algorithm for Localizing Multiple Sound Sources in Reverberant Environments An EM Algorithm for Localizing Multiple Sound Sources in Reverberant Environments Michael I. Mandel, Daniel P. W. Ellis LabROSA, Dept. of Electrical Engineering Columbia University New York, NY {mim,dpwe}@ee.columbia.edu

More information

MODal ENergy Analysis

MODal ENergy Analysis MODal ENergy Analysis Nicolas Totaro, Jean-Louis Guyader To cite this version: Nicolas Totaro, Jean-Louis Guyader. MODal ENergy Analysis. RASD, Jul 2013, Pise, Italy. 2013. HAL Id: hal-00841467

More information

A new approach of the concept of prime number

A new approach of the concept of prime number A new approach of the concept of prime number Jamel Ghannouchi To cite this version: Jamel Ghannouchi. A new approach of the concept of prime number. 4 pages. 24. HAL Id: hal-3943 https://hal.archives-ouvertes.fr/hal-3943

More information

THE BAYESIAN ABEL BOUND ON THE MEAN SQUARE ERROR

THE BAYESIAN ABEL BOUND ON THE MEAN SQUARE ERROR THE BAYESIAN ABEL BOUND ON THE MEAN SQUARE ERROR Alexandre Renaux, Philippe Forster, Pascal Larzabal, Christ Richmond To cite this version: Alexandre Renaux, Philippe Forster, Pascal Larzabal, Christ Richmond

More information

Can we reduce health inequalities? An analysis of the English strategy ( )

Can we reduce health inequalities? An analysis of the English strategy ( ) Can we reduce health inequalities? An analysis of the English strategy (1997-2010) Johan P Mackenbach To cite this version: Johan P Mackenbach. Can we reduce health inequalities? An analysis of the English

More information

Nodal and divergence-conforming boundary-element methods applied to electromagnetic scattering problems

Nodal and divergence-conforming boundary-element methods applied to electromagnetic scattering problems Nodal and divergence-conforming boundary-element methods applied to electromagnetic scattering problems M. Afonso, Joao Vasconcelos, Renato Mesquita, Christian Vollaire, Laurent Nicolas To cite this version:

More information